Recently, an improved version of simulated annealing (SA), named coupled SA (CSA), is proposed for global optimization. The CSA is characterized by a set of parallel SA processes coupled by their acceptance probabilities. However, unlike in the acceptance process, there is no coupling and thus no cooperative behavior or information exchange in the generation process of each individual SA process. Further, the CSA generates candidate solutions in a pure random sampling, thus does not utilize the information gained during the search. Differential evolution (DE) uses mutation and crossover operators to generate new candidate solutions and thus individuals or candidate solutions cooperate and compete with each other via information exchange, which enable the search for a better solution space. From an evolutionary perspective, this paper presents an evolutionary coupled simulated annealing (CSA), named CSA-DE, by combining the CSA with the differential evolution (DE). In the CSA-DE, the operators of the DE are introduced to generate candidate solutions, thus individual SAs cooperate and compete in both the generation and acceptance processes, which improves the performance of the original CSA. Simulation results on 19 benchmark test functions show that the CSA-DE is better than the CSA and DE.
展开▼
